I cannot guarantee each version is good for what you want so try each one, find what you like, use it and post stuff with it if you want...plus feel free to bid on whatever version you like.<hr />I'm working on a small AI Discord Community you can join here: <a target="_blank" rel="ugc" href="https://discord.gg/ehhMrD4PQT">https://discord.gg/ehhMrD4PQT</a><hr /><h1 id="hex-core">Hex Core</h1><ul><li>Hex Core has a Stylized 2.5D, Semirealistic, or 2D Dark Fantasy Anime look to it depending on the prompt. The style is similar to Hexbloom V1.55 as that's the whole point.</li><li>This is the main branch of Ionsyx rebranded and updated towards the style I originally intended.</li><li>Originally, V0.5 is Ionsyx V5.9 but renamed and posted here on the new account. V1.0 of this will be the final version unless someone wants this to be updated past V1.0</li></ul>I don’t recommend starting with this if you’re new to Illustrious or Danbooru in general—unless you enjoy a challenge—since poor prompting will result in poor images.Feel free to leave suggestions on what I can add here to help fix issues you may encounter. You’re also welcome to leave critiques, whether positive or negative, as long as they’re specific to this checkpoint model. I like having feedback so I can improve.You are also free to publish images made from this model. It helps people find both the model and your images. I also enjoy checking my model pages, so it’s usually a free like from me if I enjoy what you’ve posted. Feel free to bid on the model if you like it.You can tip, like, or follow the model to show support—only if you want to. I do this purely as a hobby, so tips go toward future LoRAs I create, LoRAs used behind the scenes in models, and bidding in model auctions if I have enough buzz. The original version on the Ionsyx page (V1.0) was made on a GTX 1060 3GB GPU—surprisingly. I’m also the same person (Arctenox) as this is my New Account, I can guarantee that.<h2 id="recommended-&quot;v0.5&quot;-and-&quot;v0.6&quot;-settings:">Recommended "V0.5" and "V0.6" Settings:</h2><ul><li>Steps: 20+ / Depends on the Sampler and Scheduler used.</li><li>CFG: 3.0+</li><li>Sampler: Euler / (Mostly a personal preference and starting point.)</li><li>Scheduler: Normal / Simple (Mostly a personal preference and starting point.)</li><li>Base generation for characters can be bad, so I recommend using: ADetailers / Upscalers / HiRes</li></ul> Note: You may need to generate a few times, if you don't get what you want exactly, as always.Recommended Positives - Don't have to use it if you have something better but this is a starting point:<pre><code>masterpiece, best quality</code></pre>Recommended Negatives - Don't have to use it if you have something better but this is a starting point:<pre><code>worst quality, watermark, artist sign, artist name, username, signature, logo, four fingers, six finger</code></pre>The Negative I will be personally using for V0.6 (if you want to try it out on the model) for V0.6 is:<pre><code>worst quality, low quality, bad, lowres, poorly drawn, bad anatomy, extra digits, fewer digits, bad hands, mutated hands, four fingers, six fingers, loli, shota, aged_down, watermark, username artist name, artist logo, signature, patreon logo, realistic, photorealistic</code></pre>Note: Due to how far along this is now four fingers, six fingers might not have been retained and usable anymore instead use the danbooru equivalent: fewer digits, extra digits

Hex Core [Illustrious]

This version is a model merge between Hex Core V1.0 Checkpoint and V4.0 of Anime Screenshot Merge IL. Which this used the Competitive Model Selector (Tensor Prism) from the Tensor Prism Merge Node Pack.

Files with SHA256: 9e172411f1921e738dc8fb72f20a868d798df5ef444e95f9a926ebe3aecb9be0